OpenAI, a research company that aims to create artificial intelligence that can benefit humanity, has announced the launch of Dall-E 3, a text-to-image tool that can generate images from natural language prompts. The tool uses ChatGPT, OpenAI’s popular AI chatbot, to help users refine their requests through conversations.
Dall-E 3 will be available to ChatGPT Plus and Enterprise customers in October via the API, the company said in a statement on Wednesday. Users can type in a request for an image, such as “a cat wearing a hat”, and tweak the prompt by chatting with ChatGPT, which will suggest different options and variations. For example, ChatGPT might ask “What kind of hat do you want?” or “Do you want the cat to be realistic or cartoonish?”
“DALL-E 3 can translate nuanced requests into extremely detailed and accurate images,” the company said. “It can handle complex compositions, diverse styles, and multiple concepts in a single image.”
OpenAI said the latest version of the tool will have more safeguards to prevent misuse and abuse of the technology. For instance, Dall-E 3 will limit its ability to generate violent, adult, or hateful content, and will decline requests that ask for images of a public figure by name, or those that ask for images in the style of a living artist. The company also said creators could opt out of having some or all of their work used to train future text-to-image tools.
Dall-E 3 is the latest innovation from OpenAI, which has been at the forefront of creating AI systems that can generate realistic and creative content from text. The company’s previous versions of Dall-E and ChatGPT have attracted millions of users and generated viral images and texts.
However, OpenAI also faces competition and challenges from other players in the field of text-to-image generation. Several companies, such as Alibaba’s Tongyi Wanxiang, Midjourney and Stability AI, have developed their own image-generating models that claim to be faster and more accurate than Dall-E. Moreover, there are legal and ethical issues around AI-generated images, such as the ownership and attribution of the works. A Washington D.C. court in August ruled that a work of art created by AI without any human input could not be copyrighted under U.S. law.